<p>Artificial Intelligence (AI) demonstrates substantial potential in driving economic growth; however, its role in air pollution (AP) mitigation remains ambiguous. Whereas existing studies primarily focus on AI applications in AP reduction from a technical perspective, this paper employs a wavelet-based quantile-on-quantile approach to systematically analyse the dynamic relationship between AI development and AP across different quantiles and time horizons. Using China as a case study, empirical results reveal that AI exerts a significant negative effect on AP at high quantiles in the long term, indicating that maturing AI technologies contribute to pollution reduction. Conversely, AI adoption shows a positive correlation with AP in the short and medium term. This discrepancy arises from two key factors: substantial constraints in current AI model applications for AP mitigation, and the short-term environmental trade-offs associated with AI integration in traditional manufacturing sectors. The research reveals significant temporal and quantile heterogeneity in AI’s impact on AP: AI may exacerbate AP in the short term, while becoming an effective mitigation tool in the long term. These findings provide critical insights for policymakers to balance AI’s temporal impacts on AP, accelerate technological innovation, and ultimately achieve sustainable pollution control objectives.</p>